finetuned_mnli.log

INFO:root:Namespace(accumulate=None, batch_size=32, bert_dataset='book_corpus_wiki_en_uncased', bert_model='bert_12_768_12', dev_batch_size=8, epochs=3, gpu=True, log_interval=100, lr=5e-05, max_len=80, model_parameters=None, optimizer='bertadam', output_dir='./output_dir', seed=2, task_name='MNLI', warmup_ratio=0.1)
INFO:root:BERTClassifier(
  (bert): BERTModel(
    (encoder): BERTEncoder(
      (dropout_layer): Dropout(p = 0.1, axes=())
      (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
      (transformer_cells): HybridSequential(
        (0): BERTEncoderCell(
          (dropout_layer): Dropout(p = 0.1, axes=())
          (attention_cell): MultiHeadAttentionCell(
            (_base_cell): DotProductAttentionCell(
              (_dropout_layer): Dropout(p = 0.1, axes=())
            )
            (proj_query): Dense(768 -> 768, linear)
            (proj_key): Dense(768 -> 768, linear)
            (proj_value): Dense(768 -> 768, linear)
          )
          (proj): Dense(768 -> 768, linear)
          (ffn): BERTPositionwiseFFN(
            (ffn_1): Dense(768 -> 3072, linear)
            (activation): GELU()
            (ffn_2): Dense(3072 -> 768, linear)
            (dropout_layer): Dropout(p = 0.1, axes=())
            (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
          )
          (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
        )
        (1): BERTEncoderCell(
          (dropout_layer): Dropout(p = 0.1, axes=())
          (attention_cell): MultiHeadAttentionCell(
            (_base_cell): DotProductAttentionCell(
              (_dropout_layer): Dropout(p = 0.1, axes=())
            )
            (proj_query): Dense(768 -> 768, linear)
            (proj_key): Dense(768 -> 768, linear)
            (proj_value): Dense(768 -> 768, linear)
          )
          (proj): Dense(768 -> 768, linear)
          (ffn): BERTPositionwiseFFN(
            (ffn_1): Dense(768 -> 3072, linear)
            (activation): GELU()
            (ffn_2): Dense(3072 -> 768, linear)
            (dropout_layer): Dropout(p = 0.1, axes=())
            (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
          )
          (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
        )
        (2): BERTEncoderCell(
          (dropout_layer): Dropout(p = 0.1, axes=())
          (attention_cell): MultiHeadAttentionCell(
            (_base_cell): DotProductAttentionCell(
              (_dropout_layer): Dropout(p = 0.1, axes=())
            )
            (proj_query): Dense(768 -> 768, linear)
            (proj_key): Dense(768 -> 768, linear)
            (proj_value): Dense(768 -> 768, linear)
          )
          (proj): Dense(768 -> 768, linear)
          (ffn): BERTPositionwiseFFN(
            (ffn_1): Dense(768 -> 3072, linear)
            (activation): GELU()
            (ffn_2): Dense(3072 -> 768, linear)
            (dropout_layer): Dropout(p = 0.1, axes=())
            (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
          )
          (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
        )
        (3): BERTEncoderCell(
          (dropout_layer): Dropout(p = 0.1, axes=())
          (attention_cell): MultiHeadAttentionCell(
            (_base_cell): DotProductAttentionCell(
              (_dropout_layer): Dropout(p = 0.1, axes=())
            )
            (proj_query): Dense(768 -> 768, linear)
            (proj_key): Dense(768 -> 768, linear)
            (proj_value): Dense(768 -> 768, linear)
          )
          (proj): Dense(768 -> 768, linear)
          (ffn): BERTPositionwiseFFN(
            (ffn_1): Dense(768 -> 3072, linear)
            (activation): GELU()
            (ffn_2): Dense(3072 -> 768, linear)
            (dropout_layer): Dropout(p = 0.1, axes=())
            (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
          )
          (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
        )
        (4): BERTEncoderCell(
          (dropout_layer): Dropout(p = 0.1, axes=())
          (attention_cell): MultiHeadAttentionCell(
            (_base_cell): DotProductAttentionCell(
              (_dropout_layer): Dropout(p = 0.1, axes=())
            )
            (proj_query): Dense(768 -> 768, linear)
            (proj_key): Dense(768 -> 768, linear)
            (proj_value): Dense(768 -> 768, linear)
          )
          (proj): Dense(768 -> 768, linear)
          (ffn): BERTPositionwiseFFN(
            (ffn_1): Dense(768 -> 3072, linear)
            (activation): GELU()
            (ffn_2): Dense(3072 -> 768, linear)
            (dropout_layer): Dropout(p = 0.1, axes=())
            (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
          )
          (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
        )
        (5): BERTEncoderCell(
          (dropout_layer): Dropout(p = 0.1, axes=())
          (attention_cell): MultiHeadAttentionCell(
            (_base_cell): DotProductAttentionCell(
              (_dropout_layer): Dropout(p = 0.1, axes=())
            )
            (proj_query): Dense(768 -> 768, linear)
            (proj_key): Dense(768 -> 768, linear)
            (proj_value): Dense(768 -> 768, linear)
          )
          (proj): Dense(768 -> 768, linear)
          (ffn): BERTPositionwiseFFN(
            (ffn_1): Dense(768 -> 3072, linear)
            (activation): GELU()
            (ffn_2): Dense(3072 -> 768, linear)
            (dropout_layer): Dropout(p = 0.1, axes=())
            (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
          )
          (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
        )
        (6): BERTEncoderCell(
          (dropout_layer): Dropout(p = 0.1, axes=())
          (attention_cell): MultiHeadAttentionCell(
            (_base_cell): DotProductAttentionCell(
              (_dropout_layer): Dropout(p = 0.1, axes=())
            )
            (proj_query): Dense(768 -> 768, linear)
            (proj_key): Dense(768 -> 768, linear)
            (proj_value): Dense(768 -> 768, linear)
          )
          (proj): Dense(768 -> 768, linear)
          (ffn): BERTPositionwiseFFN(
            (ffn_1): Dense(768 -> 3072, linear)
            (activation): GELU()
            (ffn_2): Dense(3072 -> 768, linear)
            (dropout_layer): Dropout(p = 0.1, axes=())
            (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
          )
          (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
        )
        (7): BERTEncoderCell(
          (dropout_layer): Dropout(p = 0.1, axes=())
          (attention_cell): MultiHeadAttentionCell(
            (_base_cell): DotProductAttentionCell(
              (_dropout_layer): Dropout(p = 0.1, axes=())
            )
            (proj_query): Dense(768 -> 768, linear)
            (proj_key): Dense(768 -> 768, linear)
            (proj_value): Dense(768 -> 768, linear)
          )
          (proj): Dense(768 -> 768, linear)
          (ffn): BERTPositionwiseFFN(
            (ffn_1): Dense(768 -> 3072, linear)
            (activation): GELU()
            (ffn_2): Dense(3072 -> 768, linear)
            (dropout_layer): Dropout(p = 0.1, axes=())
            (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
          )
          (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
        )
        (8): BERTEncoderCell(
          (dropout_layer): Dropout(p = 0.1, axes=())
          (attention_cell): MultiHeadAttentionCell(
            (_base_cell): DotProductAttentionCell(
              (_dropout_layer): Dropout(p = 0.1, axes=())
            )
            (proj_query): Dense(768 -> 768, linear)
            (proj_key): Dense(768 -> 768, linear)
            (proj_value): Dense(768 -> 768, linear)
          )
          (proj): Dense(768 -> 768, linear)
          (ffn): BERTPositionwiseFFN(
            (ffn_1): Dense(768 -> 3072, linear)
            (activation): GELU()
            (ffn_2): Dense(3072 -> 768, linear)
            (dropout_layer): Dropout(p = 0.1, axes=())
            (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
          )
          (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
        )
        (9): BERTEncoderCell(
          (dropout_layer): Dropout(p = 0.1, axes=())
          (attention_cell): MultiHeadAttentionCell(
            (_base_cell): DotProductAttentionCell(
              (_dropout_layer): Dropout(p = 0.1, axes=())
            )
            (proj_query): Dense(768 -> 768, linear)
            (proj_key): Dense(768 -> 768, linear)
            (proj_value): Dense(768 -> 768, linear)
          )
          (proj): Dense(768 -> 768, linear)
          (ffn): BERTPositionwiseFFN(
            (ffn_1): Dense(768 -> 3072, linear)
            (activation): GELU()
            (ffn_2): Dense(3072 -> 768, linear)
            (dropout_layer): Dropout(p = 0.1, axes=())
            (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
          )
          (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
        )
        (10): BERTEncoderCell(
          (dropout_layer): Dropout(p = 0.1, axes=())
          (attention_cell): MultiHeadAttentionCell(
            (_base_cell): DotProductAttentionCell(
              (_dropout_layer): Dropout(p = 0.1, axes=())
            )
            (proj_query): Dense(768 -> 768, linear)
            (proj_key): Dense(768 -> 768, linear)
            (proj_value): Dense(768 -> 768, linear)
          )
          (proj): Dense(768 -> 768, linear)
          (ffn): BERTPositionwiseFFN(
            (ffn_1): Dense(768 -> 3072, linear)
            (activation): GELU()
            (ffn_2): Dense(3072 -> 768, linear)
            (dropout_layer): Dropout(p = 0.1, axes=())
            (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
          )
          (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
        )
        (11): BERTEncoderCell(
          (dropout_layer): Dropout(p = 0.1, axes=())
          (attention_cell): MultiHeadAttentionCell(
            (_base_cell): DotProductAttentionCell(
              (_dropout_layer): Dropout(p = 0.1, axes=())
            )
            (proj_query): Dense(768 -> 768, linear)
            (proj_key): Dense(768 -> 768, linear)
            (proj_value): Dense(768 -> 768, linear)
          )
          (proj): Dense(768 -> 768, linear)
          (ffn): BERTPositionwiseFFN(
            (ffn_1): Dense(768 -> 3072, linear)
            (activation): GELU()
            (ffn_2): Dense(3072 -> 768, linear)
            (dropout_layer): Dropout(p = 0.1, axes=())
            (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
          )
          (layer_norm): BERTLayerNorm(eps=1e-12, axis=-1, center=True, scale=True, in_channels=768)
        )
      )
    )
    (word_embed): HybridSequential(
      (0): Embedding(30522 -> 768, float32)
      (1): Dropout(p = 0.1, axes=())
    )
    (token_type_embed): HybridSequential(
      (0): Embedding(2 -> 768, float32)
      (1): Dropout(p = 0.1, axes=())
    )
    (pooler): Dense(768 -> 768, Activation(tanh))
  )
  (classifier): HybridSequential(
    (0): Dropout(p = 0.1, axes=())
    (1): Dense(None -> 3, linear)
  )
)
INFO:root:processing dataset...
INFO:root:[Epoch 1 Batch 100/12277] loss=1.1486, lr=0.0000014, metrics=accuracy:0.3550
INFO:root:[Epoch 1 Batch 200/12277] loss=1.0610, lr=0.0000027, metrics=accuracy:0.3909
INFO:root:[Epoch 1 Batch 300/12277] loss=0.9557, lr=0.0000041, metrics=accuracy:0.4423
INFO:root:[Epoch 1 Batch 400/12277] loss=0.8391, lr=0.0000054, metrics=accuracy:0.4914
INFO:root:[Epoch 1 Batch 500/12277] loss=0.7997, lr=0.0000068, metrics=accuracy:0.5241
INFO:root:[Epoch 1 Batch 600/12277] loss=0.7553, lr=0.0000081, metrics=accuracy:0.5497
INFO:root:[Epoch 1 Batch 700/12277] loss=0.7370, lr=0.0000095, metrics=accuracy:0.5687
INFO:root:[Epoch 1 Batch 800/12277] loss=0.7035, lr=0.0000109, metrics=accuracy:0.5856
INFO:root:[Epoch 1 Batch 900/12277] loss=0.6920, lr=0.0000122, metrics=accuracy:0.5992
INFO:root:[Epoch 1 Batch 1000/12277] loss=0.6913, lr=0.0000136, metrics=accuracy:0.6103
INFO:root:[Epoch 1 Batch 1100/12277] loss=0.6783, lr=0.0000149, metrics=accuracy:0.6197
INFO:root:[Epoch 1 Batch 1200/12277] loss=0.6398, lr=0.0000163, metrics=accuracy:0.6297
INFO:root:[Epoch 1 Batch 1300/12277] loss=0.6366, lr=0.0000177, metrics=accuracy:0.6385
INFO:root:[Epoch 1 Batch 1400/12277] loss=0.6478, lr=0.0000190, metrics=accuracy:0.6453
INFO:root:[Epoch 1 Batch 1500/12277] loss=0.6242, lr=0.0000204, metrics=accuracy:0.6520
INFO:root:[Epoch 1 Batch 1600/12277] loss=0.6165, lr=0.0000217, metrics=accuracy:0.6583
INFO:root:[Epoch 1 Batch 1700/12277] loss=0.6019, lr=0.0000231, metrics=accuracy:0.6639
INFO:root:[Epoch 1 Batch 1800/12277] loss=0.6262, lr=0.0000244, metrics=accuracy:0.6687
INFO:root:[Epoch 1 Batch 1900/12277] loss=0.6176, lr=0.0000258, metrics=accuracy:0.6730
INFO:root:[Epoch 1 Batch 2000/12277] loss=0.5914, lr=0.0000272, metrics=accuracy:0.6771
INFO:root:[Epoch 1 Batch 2100/12277] loss=0.5881, lr=0.0000285, metrics=accuracy:0.6814
INFO:root:[Epoch 1 Batch 2200/12277] loss=0.6084, lr=0.0000299, metrics=accuracy:0.6846
INFO:root:[Epoch 1 Batch 2300/12277] loss=0.6204, lr=0.0000312, metrics=accuracy:0.6877
INFO:root:[Epoch 1 Batch 2400/12277] loss=0.5828, lr=0.0000326, metrics=accuracy:0.6909
INFO:root:[Epoch 1 Batch 2500/12277] loss=0.6147, lr=0.0000340, metrics=accuracy:0.6930
INFO:root:[Epoch 1 Batch 2600/12277] loss=0.5919, lr=0.0000353, metrics=accuracy:0.6955
INFO:root:[Epoch 1 Batch 2700/12277] loss=0.6018, lr=0.0000367, metrics=accuracy:0.6980
INFO:root:[Epoch 1 Batch 2800/12277] loss=0.6102, lr=0.0000380, metrics=accuracy:0.6999
INFO:root:[Epoch 1 Batch 2900/12277] loss=0.5564, lr=0.0000394, metrics=accuracy:0.7026
INFO:root:[Epoch 1 Batch 3000/12277] loss=0.6061, lr=0.0000407, metrics=accuracy:0.7044
INFO:root:[Epoch 1 Batch 3100/12277] loss=0.5720, lr=0.0000421, metrics=accuracy:0.7067
INFO:root:[Epoch 1 Batch 3200/12277] loss=0.5922, lr=0.0000435, metrics=accuracy:0.7081
INFO:root:[Epoch 1 Batch 3300/12277] loss=0.5786, lr=0.0000448, metrics=accuracy:0.7100
INFO:root:[Epoch 1 Batch 3400/12277] loss=0.5656, lr=0.0000462, metrics=accuracy:0.7119
INFO:root:[Epoch 1 Batch 3500/12277] loss=0.5765, lr=0.0000475, metrics=accuracy:0.7134
INFO:root:[Epoch 1 Batch 3600/12277] loss=0.5689, lr=0.0000489, metrics=accuracy:0.7150
INFO:root:[Epoch 1 Batch 3700/12277] loss=0.5286, lr=0.0000500, metrics=accuracy:0.7170
INFO:root:[Epoch 1 Batch 3800/12277] loss=0.5646, lr=0.0000498, metrics=accuracy:0.7186
INFO:root:[Epoch 1 Batch 3900/12277] loss=0.5563, lr=0.0000497, metrics=accuracy:0.7201
INFO:root:[Epoch 1 Batch 4000/12277] loss=0.5847, lr=0.0000495, metrics=accuracy:0.7212
INFO:root:[Epoch 1 Batch 4100/12277] loss=0.5873, lr=0.0000494, metrics=accuracy:0.7225
INFO:root:[Epoch 1 Batch 4200/12277] loss=0.5530, lr=0.0000492, metrics=accuracy:0.7238
INFO:root:[Epoch 1 Batch 4300/12277] loss=0.5895, lr=0.0000491, metrics=accuracy:0.7246
INFO:root:[Epoch 1 Batch 4400/12277] loss=0.5605, lr=0.0000489, metrics=accuracy:0.7258
INFO:root:[Epoch 1 Batch 4500/12277] loss=0.5573, lr=0.0000488, metrics=accuracy:0.7270
INFO:root:[Epoch 1 Batch 4600/12277] loss=0.5499, lr=0.0000486, metrics=accuracy:0.7283
INFO:root:[Epoch 1 Batch 4700/12277] loss=0.5515, lr=0.0000485, metrics=accuracy:0.7294
INFO:root:[Epoch 1 Batch 4800/12277] loss=0.5390, lr=0.0000483, metrics=accuracy:0.7308
INFO:root:[Epoch 1 Batch 4900/12277] loss=0.5654, lr=0.0000482, metrics=accuracy:0.7316
INFO:root:[Epoch 1 Batch 5000/12277] loss=0.6041, lr=0.0000480, metrics=accuracy:0.7322
INFO:root:[Epoch 1 Batch 5100/12277] loss=0.5512, lr=0.0000479, metrics=accuracy:0.7333
INFO:root:[Epoch 1 Batch 5200/12277] loss=0.5534, lr=0.0000477, metrics=accuracy:0.7341
INFO:root:[Epoch 1 Batch 5300/12277] loss=0.5583, lr=0.0000476, metrics=accuracy:0.7348
INFO:root:[Epoch 1 Batch 5400/12277] loss=0.5186, lr=0.0000474, metrics=accuracy:0.7359
INFO:root:[Epoch 1 Batch 5500/12277] loss=0.5604, lr=0.0000473, metrics=accuracy:0.7365
INFO:root:[Epoch 1 Batch 5600/12277] loss=0.5303, lr=0.0000471, metrics=accuracy:0.7373
INFO:root:[Epoch 1 Batch 5700/12277] loss=0.5473, lr=0.0000470, metrics=accuracy:0.7381
INFO:root:[Epoch 1 Batch 5800/12277] loss=0.5175, lr=0.0000468, metrics=accuracy:0.7391
INFO:root:[Epoch 1 Batch 5900/12277] loss=0.5376, lr=0.0000467, metrics=accuracy:0.7398
INFO:root:[Epoch 1 Batch 6000/12277] loss=0.5551, lr=0.0000465, metrics=accuracy:0.7404
INFO:root:[Epoch 1 Batch 6100/12277] loss=0.5492, lr=0.0000463, metrics=accuracy:0.7410
INFO:root:[Epoch 1 Batch 6200/12277] loss=0.5311, lr=0.0000462, metrics=accuracy:0.7417
INFO:root:[Epoch 1 Batch 6300/12277] loss=0.5236, lr=0.0000460, metrics=accuracy:0.7425
INFO:root:[Epoch 1 Batch 6400/12277] loss=0.5166, lr=0.0000459, metrics=accuracy:0.7433
INFO:root:[Epoch 1 Batch 6500/12277] loss=0.5445, lr=0.0000457, metrics=accuracy:0.7440
INFO:root:[Epoch 1 Batch 6600/12277] loss=0.5348, lr=0.0000456, metrics=accuracy:0.7447
INFO:root:[Epoch 1 Batch 6700/12277] loss=0.5247, lr=0.0000454, metrics=accuracy:0.7454
INFO:root:[Epoch 1 Batch 6800/12277] loss=0.5279, lr=0.0000453, metrics=accuracy:0.7462
INFO:root:[Epoch 1 Batch 6900/12277] loss=0.5058, lr=0.0000451, metrics=accuracy:0.7470
INFO:root:[Epoch 1 Batch 7000/12277] loss=0.5217, lr=0.0000450, metrics=accuracy:0.7476
INFO:root:[Epoch 1 Batch 7100/12277] loss=0.5214, lr=0.0000448, metrics=accuracy:0.7482
INFO:root:[Epoch 1 Batch 7200/12277] loss=0.5147, lr=0.0000447, metrics=accuracy:0.7489
INFO:root:[Epoch 1 Batch 7300/12277] loss=0.5265, lr=0.0000445, metrics=accuracy:0.7496
INFO:root:[Epoch 1 Batch 7400/12277] loss=0.5113, lr=0.0000444, metrics=accuracy:0.7502
INFO:root:[Epoch 1 Batch 7500/12277] loss=0.5129, lr=0.0000442, metrics=accuracy:0.7509
INFO:root:[Epoch 1 Batch 7600/12277] loss=0.5286, lr=0.0000441, metrics=accuracy:0.7514
INFO:root:[Epoch 1 Batch 7700/12277] loss=0.5334, lr=0.0000439, metrics=accuracy:0.7519
INFO:root:[Epoch 1 Batch 7800/12277] loss=0.4988, lr=0.0000438, metrics=accuracy:0.7526
INFO:root:[Epoch 1 Batch 7900/12277] loss=0.5064, lr=0.0000436, metrics=accuracy:0.7532
INFO:root:[Epoch 1 Batch 8000/12277] loss=0.4915, lr=0.0000435, metrics=accuracy:0.7538
INFO:root:[Epoch 1 Batch 8100/12277] loss=0.5063, lr=0.0000433, metrics=accuracy:0.7545
INFO:root:[Epoch 1 Batch 8200/12277] loss=0.4974, lr=0.0000432, metrics=accuracy:0.7551
INFO:root:[Epoch 1 Batch 8300/12277] loss=0.5069, lr=0.0000430, metrics=accuracy:0.7557
INFO:root:[Epoch 1 Batch 8400/12277] loss=0.5158, lr=0.0000429, metrics=accuracy:0.7561
INFO:root:[Epoch 1 Batch 8500/12277] loss=0.4869, lr=0.0000427, metrics=accuracy:0.7567
INFO:root:[Epoch 1 Batch 8600/12277] loss=0.5133, lr=0.0000426, metrics=accuracy:0.7572
INFO:root:[Epoch 1 Batch 8700/12277] loss=0.5269, lr=0.0000424, metrics=accuracy:0.7576
INFO:root:[Epoch 1 Batch 8800/12277] loss=0.5105, lr=0.0000423, metrics=accuracy:0.7580
INFO:root:[Epoch 1 Batch 8900/12277] loss=0.5210, lr=0.0000421, metrics=accuracy:0.7585
INFO:root:[Epoch 1 Batch 9000/12277] loss=0.4871, lr=0.0000420, metrics=accuracy:0.7590
INFO:root:[Epoch 1 Batch 9100/12277] loss=0.5097, lr=0.0000418, metrics=accuracy:0.7594
INFO:root:[Epoch 1 Batch 9200/12277] loss=0.5006, lr=0.0000417, metrics=accuracy:0.7599
INFO:root:[Epoch 1 Batch 9300/12277] loss=0.5186, lr=0.0000415, metrics=accuracy:0.7603
INFO:root:[Epoch 1 Batch 9400/12277] loss=0.5054, lr=0.0000414, metrics=accuracy:0.7608
INFO:root:[Epoch 1 Batch 9500/12277] loss=0.5145, lr=0.0000412, metrics=accuracy:0.7612
INFO:root:[Epoch 1 Batch 9600/12277] loss=0.5008, lr=0.0000411, metrics=accuracy:0.7616
INFO:root:[Epoch 1 Batch 9700/12277] loss=0.5175, lr=0.0000409, metrics=accuracy:0.7619
INFO:root:[Epoch 1 Batch 9800/12277] loss=0.5049, lr=0.0000408, metrics=accuracy:0.7624
INFO:root:[Epoch 1 Batch 9900/12277] loss=0.4772, lr=0.0000406, metrics=accuracy:0.7629
INFO:root:[Epoch 1 Batch 10000/12277] loss=0.4961, lr=0.0000405, metrics=accuracy:0.7633
INFO:root:[Epoch 1 Batch 10100/12277] loss=0.4888, lr=0.0000403, metrics=accuracy:0.7637
INFO:root:[Epoch 1 Batch 10200/12277] loss=0.5006, lr=0.0000402, metrics=accuracy:0.7641
INFO:root:[Epoch 1 Batch 10300/12277] loss=0.4961, lr=0.0000400, metrics=accuracy:0.7645
INFO:root:[Epoch 1 Batch 10400/12277] loss=0.4925, lr=0.0000399, metrics=accuracy:0.7649
INFO:root:[Epoch 1 Batch 10500/12277] loss=0.4924, lr=0.0000397, metrics=accuracy:0.7653
INFO:root:[Epoch 1 Batch 10600/12277] loss=0.4956, lr=0.0000396, metrics=accuracy:0.7657
INFO:root:[Epoch 1 Batch 10700/12277] loss=0.4858, lr=0.0000394, metrics=accuracy:0.7660
INFO:root:[Epoch 1 Batch 10800/12277] loss=0.5099, lr=0.0000393, metrics=accuracy:0.7663
INFO:root:[Epoch 1 Batch 10900/12277] loss=0.5028, lr=0.0000391, metrics=accuracy:0.7666
INFO:root:[Epoch 1 Batch 11000/12277] loss=0.4804, lr=0.0000390, metrics=accuracy:0.7670
INFO:root:[Epoch 1 Batch 11100/12277] loss=0.4793, lr=0.0000388, metrics=accuracy:0.7675
INFO:root:[Epoch 1 Batch 11200/12277] loss=0.4859, lr=0.0000387, metrics=accuracy:0.7679
INFO:root:[Epoch 1 Batch 11300/12277] loss=0.4899, lr=0.0000385, metrics=accuracy:0.7683
INFO:root:[Epoch 1 Batch 11400/12277] loss=0.4935, lr=0.0000384, metrics=accuracy:0.7686
INFO:root:[Epoch 1 Batch 11500/12277] loss=0.4863, lr=0.0000382, metrics=accuracy:0.7690
INFO:root:[Epoch 1 Batch 11600/12277] loss=0.4926, lr=0.0000381, metrics=accuracy:0.7693
INFO:root:[Epoch 1 Batch 11700/12277] loss=0.4809, lr=0.0000379, metrics=accuracy:0.7696
INFO:root:[Epoch 1 Batch 11800/12277] loss=0.4947, lr=0.0000377, metrics=accuracy:0.7699
INFO:root:[Epoch 1 Batch 11900/12277] loss=0.4842, lr=0.0000376, metrics=accuracy:0.7702
INFO:root:[Epoch 1 Batch 12000/12277] loss=0.4901, lr=0.0000374, metrics=accuracy:0.7705
INFO:root:[Epoch 1 Batch 12100/12277] loss=0.4987, lr=0.0000373, metrics=accuracy:0.7709
INFO:root:[Epoch 1 Batch 12200/12277] loss=0.4914, lr=0.0000371, metrics=accuracy:0.7711
INFO:root:On MNLI Matched: 
INFO:root:validation metrics:accuracy:0.8175
INFO:root:On MNLI Mismatched: 
INFO:root:validation metrics:accuracy:0.8247
INFO:root:params saved in : ./output_dir/model_bert_MNLI_0.params
INFO:root:Time cost=2930.7s
INFO:root:[Epoch 2 Batch 100/12277] loss=0.3645, lr=0.0000369, metrics=accuracy:0.8569
INFO:root:[Epoch 2 Batch 200/12277] loss=0.3640, lr=0.0000367, metrics=accuracy:0.8609
INFO:root:[Epoch 2 Batch 300/12277] loss=0.3761, lr=0.0000366, metrics=accuracy:0.8609
INFO:root:[Epoch 2 Batch 400/12277] loss=0.3780, lr=0.0000364, metrics=accuracy:0.8606
INFO:root:[Epoch 2 Batch 500/12277] loss=0.3951, lr=0.0000363, metrics=accuracy:0.8580
INFO:root:[Epoch 2 Batch 600/12277] loss=0.3729, lr=0.0000361, metrics=accuracy:0.8578
INFO:root:[Epoch 2 Batch 700/12277] loss=0.3841, lr=0.0000360, metrics=accuracy:0.8572
INFO:root:[Epoch 2 Batch 800/12277] loss=0.3800, lr=0.0000358, metrics=accuracy:0.8573
INFO:root:[Epoch 2 Batch 900/12277] loss=0.4055, lr=0.0000357, metrics=accuracy:0.8557
INFO:root:[Epoch 2 Batch 1000/12277] loss=0.3628, lr=0.0000355, metrics=accuracy:0.8563
INFO:root:[Epoch 2 Batch 1100/12277] loss=0.3732, lr=0.0000354, metrics=accuracy:0.8561
INFO:root:[Epoch 2 Batch 1200/12277] loss=0.3845, lr=0.0000352, metrics=accuracy:0.8555
INFO:root:[Epoch 2 Batch 1300/12277] loss=0.3997, lr=0.0000351, metrics=accuracy:0.8552
INFO:root:[Epoch 2 Batch 1400/12277] loss=0.3712, lr=0.0000349, metrics=accuracy:0.8557
INFO:root:[Epoch 2 Batch 1500/12277] loss=0.3610, lr=0.0000348, metrics=accuracy:0.8563
INFO:root:[Epoch 2 Batch 1600/12277] loss=0.3803, lr=0.0000346, metrics=accuracy:0.8564
INFO:root:[Epoch 2 Batch 1700/12277] loss=0.3863, lr=0.0000345, metrics=accuracy:0.8560
INFO:root:[Epoch 2 Batch 1800/12277] loss=0.3986, lr=0.0000343, metrics=accuracy:0.8556
INFO:root:[Epoch 2 Batch 1900/12277] loss=0.3859, lr=0.0000342, metrics=accuracy:0.8557
INFO:root:[Epoch 2 Batch 2000/12277] loss=0.3711, lr=0.0000340, metrics=accuracy:0.8557
INFO:root:[Epoch 2 Batch 2100/12277] loss=0.3509, lr=0.0000339, metrics=accuracy:0.8564
INFO:root:[Epoch 2 Batch 2200/12277] loss=0.3872, lr=0.0000337, metrics=accuracy:0.8561
INFO:root:[Epoch 2 Batch 2300/12277] loss=0.3888, lr=0.0000336, metrics=accuracy:0.8560
INFO:root:[Epoch 2 Batch 2400/12277] loss=0.3621, lr=0.0000334, metrics=accuracy:0.8562
INFO:root:[Epoch 2 Batch 2500/12277] loss=0.4020, lr=0.0000333, metrics=accuracy:0.8559
INFO:root:[Epoch 2 Batch 2600/12277] loss=0.3661, lr=0.0000331, metrics=accuracy:0.8560
INFO:root:[Epoch 2 Batch 2700/12277] loss=0.3716, lr=0.0000330, metrics=accuracy:0.8563
INFO:root:[Epoch 2 Batch 2800/12277] loss=0.3582, lr=0.0000328, metrics=accuracy:0.8564
INFO:root:[Epoch 2 Batch 2900/12277] loss=0.3753, lr=0.0000327, metrics=accuracy:0.8566
INFO:root:[Epoch 2 Batch 3000/12277] loss=0.3692, lr=0.0000325, metrics=accuracy:0.8569
INFO:root:[Epoch 2 Batch 3100/12277] loss=0.3805, lr=0.0000324, metrics=accuracy:0.8568
INFO:root:[Epoch 2 Batch 3200/12277] loss=0.4049, lr=0.0000322, metrics=accuracy:0.8565
INFO:root:[Epoch 2 Batch 3300/12277] loss=0.3686, lr=0.0000320, metrics=accuracy:0.8567
INFO:root:[Epoch 2 Batch 3400/12277] loss=0.3758, lr=0.0000319, metrics=accuracy:0.8567
INFO:root:[Epoch 2 Batch 3500/12277] loss=0.3626, lr=0.0000317, metrics=accuracy:0.8569
INFO:root:[Epoch 2 Batch 3600/12277] loss=0.3679, lr=0.0000316, metrics=accuracy:0.8570
INFO:root:[Epoch 2 Batch 3700/12277] loss=0.3606, lr=0.0000314, metrics=accuracy:0.8571
INFO:root:[Epoch 2 Batch 3800/12277] loss=0.3941, lr=0.0000313, metrics=accuracy:0.8569
INFO:root:[Epoch 2 Batch 3900/12277] loss=0.3638, lr=0.0000311, metrics=accuracy:0.8570
INFO:root:[Epoch 2 Batch 4000/12277] loss=0.3755, lr=0.0000310, metrics=accuracy:0.8570
INFO:root:[Epoch 2 Batch 4100/12277] loss=0.3819, lr=0.0000308, metrics=accuracy:0.8570
INFO:root:[Epoch 2 Batch 4200/12277] loss=0.3784, lr=0.0000307, metrics=accuracy:0.8569
INFO:root:[Epoch 2 Batch 4300/12277] loss=0.3730, lr=0.0000305, metrics=accuracy:0.8570
INFO:root:[Epoch 2 Batch 4400/12277] loss=0.3714, lr=0.0000304, metrics=accuracy:0.8571
INFO:root:[Epoch 2 Batch 4500/12277] loss=0.3925, lr=0.0000302, metrics=accuracy:0.8570
INFO:root:[Epoch 2 Batch 4600/12277] loss=0.3852, lr=0.0000301, metrics=accuracy:0.8570
INFO:root:[Epoch 2 Batch 4700/12277] loss=0.3580, lr=0.0000299, metrics=accuracy:0.8573
INFO:root:[Epoch 2 Batch 4800/12277] loss=0.3778, lr=0.0000298, metrics=accuracy:0.8572
INFO:root:[Epoch 2 Batch 4900/12277] loss=0.3706, lr=0.0000296, metrics=accuracy:0.8572
INFO:root:[Epoch 2 Batch 5000/12277] loss=0.3776, lr=0.0000295, metrics=accuracy:0.8573
INFO:root:[Epoch 2 Batch 5100/12277] loss=0.3687, lr=0.0000293, metrics=accuracy:0.8573
INFO:root:[Epoch 2 Batch 5200/12277] loss=0.3880, lr=0.0000292, metrics=accuracy:0.8573
INFO:root:[Epoch 2 Batch 5300/12277] loss=0.3875, lr=0.0000290, metrics=accuracy:0.8571
INFO:root:[Epoch 2 Batch 5400/12277] loss=0.3697, lr=0.0000289, metrics=accuracy:0.8572
INFO:root:[Epoch 2 Batch 5500/12277] loss=0.3629, lr=0.0000287, metrics=accuracy:0.8572
INFO:root:[Epoch 2 Batch 5600/12277] loss=0.3942, lr=0.0000286, metrics=accuracy:0.8572
INFO:root:[Epoch 2 Batch 5700/12277] loss=0.3857, lr=0.0000284, metrics=accuracy:0.8571
INFO:root:[Epoch 2 Batch 5800/12277] loss=0.3791, lr=0.0000283, metrics=accuracy:0.8571
INFO:root:[Epoch 2 Batch 5900/12277] loss=0.3851, lr=0.0000281, metrics=accuracy:0.8571
INFO:root:[Epoch 2 Batch 6000/12277] loss=0.3544, lr=0.0000280, metrics=accuracy:0.8571
INFO:root:[Epoch 2 Batch 6100/12277] loss=0.3752, lr=0.0000278, metrics=accuracy:0.8571
INFO:root:[Epoch 2 Batch 6200/12277] loss=0.3743, lr=0.0000277, metrics=accuracy:0.8571
INFO:root:[Epoch 2 Batch 6300/12277] loss=0.3695, lr=0.0000275, metrics=accuracy:0.8572
INFO:root:[Epoch 2 Batch 6400/12277] loss=0.3647, lr=0.0000274, metrics=accuracy:0.8573
INFO:root:[Epoch 2 Batch 6500/12277] loss=0.3683, lr=0.0000272, metrics=accuracy:0.8574
INFO:root:[Epoch 2 Batch 6600/12277] loss=0.3642, lr=0.0000271, metrics=accuracy:0.8574
INFO:root:[Epoch 2 Batch 6700/12277] loss=0.3791, lr=0.0000269, metrics=accuracy:0.8574
INFO:root:[Epoch 2 Batch 6800/12277] loss=0.3630, lr=0.0000268, metrics=accuracy:0.8573
INFO:root:[Epoch 2 Batch 6900/12277] loss=0.3672, lr=0.0000266, metrics=accuracy:0.8574
INFO:root:[Epoch 2 Batch 7000/12277] loss=0.3749, lr=0.0000265, metrics=accuracy:0.8574
INFO:root:[Epoch 2 Batch 7100/12277] loss=0.3806, lr=0.0000263, metrics=accuracy:0.8573
INFO:root:[Epoch 2 Batch 7200/12277] loss=0.3699, lr=0.0000262, metrics=accuracy:0.8574
INFO:root:[Epoch 2 Batch 7300/12277] loss=0.3898, lr=0.0000260, metrics=accuracy:0.8573
INFO:root:[Epoch 2 Batch 7400/12277] loss=0.3465, lr=0.0000259, metrics=accuracy:0.8574
INFO:root:[Epoch 2 Batch 7500/12277] loss=0.3575, lr=0.0000257, metrics=accuracy:0.8576
INFO:root:[Epoch 2 Batch 7600/12277] loss=0.3430, lr=0.0000256, metrics=accuracy:0.8577
INFO:root:[Epoch 2 Batch 7700/12277] loss=0.3574, lr=0.0000254, metrics=accuracy:0.8577
INFO:root:[Epoch 2 Batch 7800/12277] loss=0.3755, lr=0.0000253, metrics=accuracy:0.8577
INFO:root:[Epoch 2 Batch 7900/12277] loss=0.3786, lr=0.0000251, metrics=accuracy:0.8576
INFO:root:[Epoch 2 Batch 8000/12277] loss=0.3562, lr=0.0000250, metrics=accuracy:0.8577
INFO:root:[Epoch 2 Batch 8100/12277] loss=0.3847, lr=0.0000248, metrics=accuracy:0.8577
INFO:root:[Epoch 2 Batch 8200/12277] loss=0.3748, lr=0.0000247, metrics=accuracy:0.8576
INFO:root:[Epoch 2 Batch 8300/12277] loss=0.3725, lr=0.0000245, metrics=accuracy:0.8577
INFO:root:[Epoch 2 Batch 8400/12277] loss=0.3467, lr=0.0000244, metrics=accuracy:0.8577
INFO:root:[Epoch 2 Batch 8500/12277] loss=0.3671, lr=0.0000242, metrics=accuracy:0.8578
INFO:root:[Epoch 2 Batch 8600/12277] loss=0.3636, lr=0.0000241, metrics=accuracy:0.8578
INFO:root:[Epoch 2 Batch 8700/12277] loss=0.3905, lr=0.0000239, metrics=accuracy:0.8577
INFO:root:[Epoch 2 Batch 8800/12277] loss=0.3660, lr=0.0000237, metrics=accuracy:0.8578
INFO:root:[Epoch 2 Batch 8900/12277] loss=0.3508, lr=0.0000236, metrics=accuracy:0.8580
INFO:root:[Epoch 2 Batch 9000/12277] loss=0.3482, lr=0.0000234, metrics=accuracy:0.8581
INFO:root:[Epoch 2 Batch 9100/12277] loss=0.3469, lr=0.0000233, metrics=accuracy:0.8582
INFO:root:[Epoch 2 Batch 9200/12277] loss=0.3583, lr=0.0000231, metrics=accuracy:0.8583
INFO:root:[Epoch 2 Batch 9300/12277] loss=0.3864, lr=0.0000230, metrics=accuracy:0.8583
INFO:root:[Epoch 2 Batch 9400/12277] loss=0.3533, lr=0.0000228, metrics=accuracy:0.8584
INFO:root:[Epoch 2 Batch 9500/12277] loss=0.3712, lr=0.0000227, metrics=accuracy:0.8584
INFO:root:[Epoch 2 Batch 9600/12277] loss=0.3501, lr=0.0000225, metrics=accuracy:0.8585
INFO:root:[Epoch 2 Batch 9700/12277] loss=0.3579, lr=0.0000224, metrics=accuracy:0.8586
INFO:root:[Epoch 2 Batch 9800/12277] loss=0.3552, lr=0.0000222, metrics=accuracy:0.8586
INFO:root:[Epoch 2 Batch 9900/12277] loss=0.3855, lr=0.0000221, metrics=accuracy:0.8586
INFO:root:[Epoch 2 Batch 10000/12277] loss=0.3719, lr=0.0000219, metrics=accuracy:0.8585
INFO:root:[Epoch 2 Batch 10100/12277] loss=0.3641, lr=0.0000218, metrics=accuracy:0.8586
INFO:root:[Epoch 2 Batch 10200/12277] loss=0.3679, lr=0.0000216, metrics=accuracy:0.8586
INFO:root:[Epoch 2 Batch 10300/12277] loss=0.3640, lr=0.0000215, metrics=accuracy:0.8586
INFO:root:[Epoch 2 Batch 10400/12277] loss=0.3541, lr=0.0000213, metrics=accuracy:0.8587
INFO:root:[Epoch 2 Batch 10500/12277] loss=0.3520, lr=0.0000212, metrics=accuracy:0.8588
INFO:root:[Epoch 2 Batch 10600/12277] loss=0.3591, lr=0.0000210, metrics=accuracy:0.8588
INFO:root:[Epoch 2 Batch 10700/12277] loss=0.3589, lr=0.0000209, metrics=accuracy:0.8588
INFO:root:[Epoch 2 Batch 10800/12277] loss=0.3673, lr=0.0000207, metrics=accuracy:0.8589
INFO:root:[Epoch 2 Batch 10900/12277] loss=0.3600, lr=0.0000206, metrics=accuracy:0.8589
INFO:root:[Epoch 2 Batch 11000/12277] loss=0.3375, lr=0.0000204, metrics=accuracy:0.8590
INFO:root:[Epoch 2 Batch 11100/12277] loss=0.3672, lr=0.0000203, metrics=accuracy:0.8590
INFO:root:[Epoch 2 Batch 11200/12277] loss=0.3659, lr=0.0000201, metrics=accuracy:0.8590
INFO:root:[Epoch 2 Batch 11300/12277] loss=0.3571, lr=0.0000200, metrics=accuracy:0.8591
INFO:root:[Epoch 2 Batch 11400/12277] loss=0.3400, lr=0.0000198, metrics=accuracy:0.8592
INFO:root:[Epoch 2 Batch 11500/12277] loss=0.3765, lr=0.0000197, metrics=accuracy:0.8591
INFO:root:[Epoch 2 Batch 11600/12277] loss=0.3333, lr=0.0000195, metrics=accuracy:0.8593
INFO:root:[Epoch 2 Batch 11700/12277] loss=0.3653, lr=0.0000194, metrics=accuracy:0.8593
INFO:root:[Epoch 2 Batch 11800/12277] loss=0.3681, lr=0.0000192, metrics=accuracy:0.8593
INFO:root:[Epoch 2 Batch 11900/12277] loss=0.3489, lr=0.0000191, metrics=accuracy:0.8595
INFO:root:[Epoch 2 Batch 12000/12277] loss=0.3388, lr=0.0000189, metrics=accuracy:0.8596
INFO:root:[Epoch 2 Batch 12100/12277] loss=0.3524, lr=0.0000188, metrics=accuracy:0.8596
INFO:root:[Epoch 2 Batch 12200/12277] loss=0.3570, lr=0.0000186, metrics=accuracy:0.8597
INFO:root:On MNLI Matched: 
INFO:root:validation metrics:accuracy:0.8395
INFO:root:On MNLI Mismatched: 
INFO:root:validation metrics:accuracy:0.8382
INFO:root:params saved in : ./output_dir/model_bert_MNLI_1.params
INFO:root:Time cost=2935.2s
INFO:root:[Epoch 3 Batch 100/12277] loss=0.2328, lr=0.0000184, metrics=accuracy:0.9181
INFO:root:[Epoch 3 Batch 200/12277] loss=0.2344, lr=0.0000182, metrics=accuracy:0.9175
INFO:root:[Epoch 3 Batch 300/12277] loss=0.2430, lr=0.0000180, metrics=accuracy:0.9172
INFO:root:[Epoch 3 Batch 400/12277] loss=0.2280, lr=0.0000179, metrics=accuracy:0.9177
INFO:root:[Epoch 3 Batch 500/12277] loss=0.2250, lr=0.0000177, metrics=accuracy:0.9183
INFO:root:[Epoch 3 Batch 600/12277] loss=0.2307, lr=0.0000176, metrics=accuracy:0.9185
INFO:root:[Epoch 3 Batch 700/12277] loss=0.2347, lr=0.0000174, metrics=accuracy:0.9181
INFO:root:[Epoch 3 Batch 800/12277] loss=0.2110, lr=0.0000173, metrics=accuracy:0.9182
INFO:root:[Epoch 3 Batch 900/12277] loss=0.2155, lr=0.0000171, metrics=accuracy:0.9189
INFO:root:[Epoch 3 Batch 1000/12277] loss=0.2089, lr=0.0000170, metrics=accuracy:0.9197
INFO:root:[Epoch 3 Batch 1100/12277] loss=0.2557, lr=0.0000168, metrics=accuracy:0.9191
INFO:root:[Epoch 3 Batch 1200/12277] loss=0.2261, lr=0.0000167, metrics=accuracy:0.9191
INFO:root:[Epoch 3 Batch 1300/12277] loss=0.2287, lr=0.0000165, metrics=accuracy:0.9191
INFO:root:[Epoch 3 Batch 1400/12277] loss=0.2201, lr=0.0000164, metrics=accuracy:0.9196
INFO:root:[Epoch 3 Batch 1500/12277] loss=0.2453, lr=0.0000162, metrics=accuracy:0.9190
INFO:root:[Epoch 3 Batch 1600/12277] loss=0.2214, lr=0.0000161, metrics=accuracy:0.9192
INFO:root:[Epoch 3 Batch 1700/12277] loss=0.2198, lr=0.0000159, metrics=accuracy:0.9191
INFO:root:[Epoch 3 Batch 1800/12277] loss=0.2280, lr=0.0000158, metrics=accuracy:0.9189
INFO:root:[Epoch 3 Batch 1900/12277] loss=0.2238, lr=0.0000156, metrics=accuracy:0.9190
INFO:root:[Epoch 3 Batch 2000/12277] loss=0.2208, lr=0.0000155, metrics=accuracy:0.9191
INFO:root:[Epoch 3 Batch 2100/12277] loss=0.2457, lr=0.0000153, metrics=accuracy:0.9187
INFO:root:[Epoch 3 Batch 2200/12277] loss=0.2342, lr=0.0000152, metrics=accuracy:0.9185
INFO:root:[Epoch 3 Batch 2300/12277] loss=0.2239, lr=0.0000150, metrics=accuracy:0.9188
INFO:root:[Epoch 3 Batch 2400/12277] loss=0.2384, lr=0.0000149, metrics=accuracy:0.9186
INFO:root:[Epoch 3 Batch 2500/12277] loss=0.2200, lr=0.0000147, metrics=accuracy:0.9188
INFO:root:[Epoch 3 Batch 2600/12277] loss=0.2378, lr=0.0000146, metrics=accuracy:0.9186
INFO:root:[Epoch 3 Batch 2700/12277] loss=0.2300, lr=0.0000144, metrics=accuracy:0.9186
INFO:root:[Epoch 3 Batch 2800/12277] loss=0.2305, lr=0.0000143, metrics=accuracy:0.9185
INFO:root:[Epoch 3 Batch 2900/12277] loss=0.2272, lr=0.0000141, metrics=accuracy:0.9186
INFO:root:[Epoch 3 Batch 3000/12277] loss=0.2408, lr=0.0000140, metrics=accuracy:0.9185
INFO:root:[Epoch 3 Batch 3100/12277] loss=0.2409, lr=0.0000138, metrics=accuracy:0.9185
INFO:root:[Epoch 3 Batch 3200/12277] loss=0.2339, lr=0.0000137, metrics=accuracy:0.9184
INFO:root:[Epoch 3 Batch 3300/12277] loss=0.2373, lr=0.0000135, metrics=accuracy:0.9183
INFO:root:[Epoch 3 Batch 3400/12277] loss=0.2138, lr=0.0000134, metrics=accuracy:0.9184
INFO:root:[Epoch 3 Batch 3500/12277] loss=0.2237, lr=0.0000132, metrics=accuracy:0.9186
INFO:root:[Epoch 3 Batch 3600/12277] loss=0.2299, lr=0.0000131, metrics=accuracy:0.9185
INFO:root:[Epoch 3 Batch 3700/12277] loss=0.2241, lr=0.0000129, metrics=accuracy:0.9185
INFO:root:[Epoch 3 Batch 3800/12277] loss=0.2216, lr=0.0000128, metrics=accuracy:0.9187
INFO:root:[Epoch 3 Batch 3900/12277] loss=0.2292, lr=0.0000126, metrics=accuracy:0.9186
INFO:root:[Epoch 3 Batch 4000/12277] loss=0.2251, lr=0.0000125, metrics=accuracy:0.9186
INFO:root:[Epoch 3 Batch 4100/12277] loss=0.2345, lr=0.0000123, metrics=accuracy:0.9184
INFO:root:[Epoch 3 Batch 4200/12277] loss=0.2460, lr=0.0000122, metrics=accuracy:0.9183
INFO:root:[Epoch 3 Batch 4300/12277] loss=0.2334, lr=0.0000120, metrics=accuracy:0.9183
INFO:root:[Epoch 3 Batch 4400/12277] loss=0.2179, lr=0.0000119, metrics=accuracy:0.9183
INFO:root:[Epoch 3 Batch 4500/12277] loss=0.2070, lr=0.0000117, metrics=accuracy:0.9186
INFO:root:[Epoch 3 Batch 4600/12277] loss=0.2398, lr=0.0000116, metrics=accuracy:0.9185
INFO:root:[Epoch 3 Batch 4700/12277] loss=0.2377, lr=0.0000114, metrics=accuracy:0.9184
INFO:root:[Epoch 3 Batch 4800/12277] loss=0.2102, lr=0.0000113, metrics=accuracy:0.9186
INFO:root:[Epoch 3 Batch 4900/12277] loss=0.2158, lr=0.0000111, metrics=accuracy:0.9187
INFO:root:[Epoch 3 Batch 5000/12277] loss=0.2282, lr=0.0000110, metrics=accuracy:0.9188
INFO:root:[Epoch 3 Batch 5100/12277] loss=0.2310, lr=0.0000108, metrics=accuracy:0.9188
INFO:root:[Epoch 3 Batch 5200/12277] loss=0.2311, lr=0.0000107, metrics=accuracy:0.9187
INFO:root:[Epoch 3 Batch 5300/12277] loss=0.2278, lr=0.0000105, metrics=accuracy:0.9188
INFO:root:[Epoch 3 Batch 5400/12277] loss=0.2296, lr=0.0000104, metrics=accuracy:0.9188
INFO:root:[Epoch 3 Batch 5500/12277] loss=0.2295, lr=0.0000102, metrics=accuracy:0.9189
INFO:root:[Epoch 3 Batch 5600/12277] loss=0.2081, lr=0.0000101, metrics=accuracy:0.9191
INFO:root:[Epoch 3 Batch 5700/12277] loss=0.2529, lr=0.0000099, metrics=accuracy:0.9190
INFO:root:[Epoch 3 Batch 5800/12277] loss=0.2334, lr=0.0000097, metrics=accuracy:0.9189
INFO:root:[Epoch 3 Batch 5900/12277] loss=0.2411, lr=0.0000096, metrics=accuracy:0.9189
INFO:root:[Epoch 3 Batch 6000/12277] loss=0.1987, lr=0.0000094, metrics=accuracy:0.9190
INFO:root:[Epoch 3 Batch 6100/12277] loss=0.2188, lr=0.0000093, metrics=accuracy:0.9191
INFO:root:[Epoch 3 Batch 6200/12277] loss=0.2191, lr=0.0000091, metrics=accuracy:0.9192
INFO:root:[Epoch 3 Batch 6300/12277] loss=0.2103, lr=0.0000090, metrics=accuracy:0.9193
INFO:root:[Epoch 3 Batch 6400/12277] loss=0.2450, lr=0.0000088, metrics=accuracy:0.9192
INFO:root:[Epoch 3 Batch 6500/12277] loss=0.2293, lr=0.0000087, metrics=accuracy:0.9191
INFO:root:[Epoch 3 Batch 6600/12277] loss=0.2226, lr=0.0000085, metrics=accuracy:0.9192
INFO:root:[Epoch 3 Batch 6700/12277] loss=0.2109, lr=0.0000084, metrics=accuracy:0.9192
INFO:root:[Epoch 3 Batch 6800/12277] loss=0.2266, lr=0.0000082, metrics=accuracy:0.9192
INFO:root:[Epoch 3 Batch 6900/12277] loss=0.2132, lr=0.0000081, metrics=accuracy:0.9193
INFO:root:[Epoch 3 Batch 7000/12277] loss=0.1935, lr=0.0000079, metrics=accuracy:0.9196
INFO:root:[Epoch 3 Batch 7100/12277] loss=0.2177, lr=0.0000078, metrics=accuracy:0.9196
INFO:root:[Epoch 3 Batch 7200/12277] loss=0.2292, lr=0.0000076, metrics=accuracy:0.9196
INFO:root:[Epoch 3 Batch 7300/12277] loss=0.2168, lr=0.0000075, metrics=accuracy:0.9196
INFO:root:[Epoch 3 Batch 7400/12277] loss=0.2034, lr=0.0000073, metrics=accuracy:0.9197
INFO:root:[Epoch 3 Batch 7500/12277] loss=0.2171, lr=0.0000072, metrics=accuracy:0.9197
INFO:root:[Epoch 3 Batch 7600/12277] loss=0.2171, lr=0.0000070, metrics=accuracy:0.9197
INFO:root:[Epoch 3 Batch 7700/12277] loss=0.2264, lr=0.0000069, metrics=accuracy:0.9197
INFO:root:[Epoch 3 Batch 7800/12277] loss=0.2149, lr=0.0000067, metrics=accuracy:0.9198
INFO:root:[Epoch 3 Batch 7900/12277] loss=0.2406, lr=0.0000066, metrics=accuracy:0.9197
INFO:root:[Epoch 3 Batch 8000/12277] loss=0.2121, lr=0.0000064, metrics=accuracy:0.9198
INFO:root:[Epoch 3 Batch 8100/12277] loss=0.2215, lr=0.0000063, metrics=accuracy:0.9198
INFO:root:[Epoch 3 Batch 8200/12277] loss=0.2263, lr=0.0000061, metrics=accuracy:0.9198
INFO:root:[Epoch 3 Batch 8300/12277] loss=0.2084, lr=0.0000060, metrics=accuracy:0.9199
INFO:root:[Epoch 3 Batch 8400/12277] loss=0.2263, lr=0.0000058, metrics=accuracy:0.9199
INFO:root:[Epoch 3 Batch 8500/12277] loss=0.2073, lr=0.0000057, metrics=accuracy:0.9200
INFO:root:[Epoch 3 Batch 8600/12277] loss=0.2363, lr=0.0000055, metrics=accuracy:0.9198
INFO:root:[Epoch 3 Batch 8700/12277] loss=0.2210, lr=0.0000054, metrics=accuracy:0.9199
INFO:root:[Epoch 3 Batch 8800/12277] loss=0.2073, lr=0.0000052, metrics=accuracy:0.9199
INFO:root:[Epoch 3 Batch 8900/12277] loss=0.2272, lr=0.0000051, metrics=accuracy:0.9199
INFO:root:[Epoch 3 Batch 9000/12277] loss=0.1974, lr=0.0000049, metrics=accuracy:0.9201
INFO:root:[Epoch 3 Batch 9100/12277] loss=0.1935, lr=0.0000048, metrics=accuracy:0.9202
INFO:root:[Epoch 3 Batch 9200/12277] loss=0.2173, lr=0.0000046, metrics=accuracy:0.9202
INFO:root:[Epoch 3 Batch 9300/12277] loss=0.2246, lr=0.0000045, metrics=accuracy:0.9203
INFO:root:[Epoch 3 Batch 9400/12277] loss=0.2251, lr=0.0000043, metrics=accuracy:0.9203
INFO:root:[Epoch 3 Batch 9500/12277] loss=0.1962, lr=0.0000042, metrics=accuracy:0.9204
INFO:root:[Epoch 3 Batch 9600/12277] loss=0.2124, lr=0.0000040, metrics=accuracy:0.9205
INFO:root:[Epoch 3 Batch 9700/12277] loss=0.2150, lr=0.0000039, metrics=accuracy:0.9205
INFO:root:[Epoch 3 Batch 9800/12277] loss=0.2148, lr=0.0000037, metrics=accuracy:0.9205
INFO:root:[Epoch 3 Batch 9900/12277] loss=0.2296, lr=0.0000036, metrics=accuracy:0.9204
INFO:root:[Epoch 3 Batch 10000/12277] loss=0.2131, lr=0.0000034, metrics=accuracy:0.9205
INFO:root:[Epoch 3 Batch 10100/12277] loss=0.2276, lr=0.0000033, metrics=accuracy:0.9205
INFO:root:[Epoch 3 Batch 10200/12277] loss=0.2342, lr=0.0000031, metrics=accuracy:0.9204
INFO:root:[Epoch 3 Batch 10300/12277] loss=0.2104, lr=0.0000030, metrics=accuracy:0.9205
INFO:root:[Epoch 3 Batch 10400/12277] loss=0.2103, lr=0.0000028, metrics=accuracy:0.9205
INFO:root:[Epoch 3 Batch 10500/12277] loss=0.2147, lr=0.0000027, metrics=accuracy:0.9206
INFO:root:[Epoch 3 Batch 10600/12277] loss=0.2185, lr=0.0000025, metrics=accuracy:0.9206
INFO:root:[Epoch 3 Batch 10700/12277] loss=0.2136, lr=0.0000024, metrics=accuracy:0.9206
INFO:root:[Epoch 3 Batch 10800/12277] loss=0.2129, lr=0.0000022, metrics=accuracy:0.9206
INFO:root:[Epoch 3 Batch 10900/12277] loss=0.2212, lr=0.0000021, metrics=accuracy:0.9206
INFO:root:[Epoch 3 Batch 11000/12277] loss=0.2069, lr=0.0000019, metrics=accuracy:0.9207
INFO:root:[Epoch 3 Batch 11100/12277] loss=0.2302, lr=0.0000018, metrics=accuracy:0.9207
INFO:root:[Epoch 3 Batch 11200/12277] loss=0.2160, lr=0.0000016, metrics=accuracy:0.9207
INFO:root:[Epoch 3 Batch 11300/12277] loss=0.1919, lr=0.0000015, metrics=accuracy:0.9207
INFO:root:[Epoch 3 Batch 11400/12277] loss=0.2249, lr=0.0000013, metrics=accuracy:0.9207
INFO:root:[Epoch 3 Batch 11500/12277] loss=0.1948, lr=0.0000011, metrics=accuracy:0.9208
INFO:root:[Epoch 3 Batch 11600/12277] loss=0.2044, lr=0.0000010, metrics=accuracy:0.9209
INFO:root:[Epoch 3 Batch 11700/12277] loss=0.1999, lr=0.0000008, metrics=accuracy:0.9209
INFO:root:[Epoch 3 Batch 11800/12277] loss=0.1921, lr=0.0000007, metrics=accuracy:0.9210
INFO:root:[Epoch 3 Batch 11900/12277] loss=0.2376, lr=0.0000005, metrics=accuracy:0.9210
INFO:root:[Epoch 3 Batch 12000/12277] loss=0.2224, lr=0.0000004, metrics=accuracy:0.9209
INFO:root:[Epoch 3 Batch 12100/12277] loss=0.2182, lr=0.0000002, metrics=accuracy:0.9210
INFO:root:[Epoch 3 Batch 12200/12277] loss=0.2330, lr=0.0000001, metrics=accuracy:0.9209
INFO:root:On MNLI Matched: 
INFO:root:validation metrics:accuracy:0.8455
INFO:root:On MNLI Mismatched: 
INFO:root:validation metrics:accuracy:0.8466
INFO:root:params saved in : ./output_dir/model_bert_MNLI_2.params
INFO:root:Time cost=2935.0s